The max aperture certainly matters in the case of DSLR PDAF if the lenses exhibit spherical aberration. And most really fast lenses have noticeable SA.
Autofocus accuracy ought to be the same - F or Z mount, fast or slow lens. A good test would be to mount the camera on a tripod and use flash exposure. That will eliminate any focus distance errors and motion blur.
Your last four words are the clue to the other significant mirrorless advantage after on-sensor PDAF.
DSLR implementations that focus only at wide open aperture are not directly dealing with focus shift, if they are at all*. The Z bodies on the other hand focus at shooting aperture up through f/5.6, after which the majority of focus shift will have already occurred.
I wonder if any DSLRs have models of the required offsets and store them in lookup tables in their lens calibration files?
sorry for coming late with an answer to your question:
I wish to emphasize that there are 3 types of error that contribute to AF error and can be distinguished, evaluated and are illustrated in my graph:
a) systematic error, can be corrected with AF tune. Is practically negligible in Z cameras.
b) hysteresis, a front back systematic error. Hopefully small enough to be negligible. Not always negligible. A minor gradual progress with later technology.
c) statistical error. Is not reproducible in focus operations. A minor gradual progress with later technology.
DSLR do AF fully open when using the OVF. This can lead to systematic AF error due to curvature of field fsor a stopped down image.
Electronic viewfinder cameras like Z do on sensor Af stopped down. This can lead to more statistical error due to noise in poor lighting.
One issue with Nikon Z focusing (and maybe other manufacturers) is that they use the post-EVF stream for focusing. That means a darker EVF has more trouble acquiring focus than a brighter EVF (brightness at the focus point matters). Some manufacturers (Ricoh) brighten the EVF quickly during focus acquisition if the EVF is too dark. Turning off Nikon's live view exposure simulation can improve focus acquisition.
Hi,
One aspect not touched upon in the replies is that the F lenses were designed pre-Z mount, and so may have a less precise / efficient electronic interface.
Additionally the Z lenses use stepping motors presumably for more precise auto-focus control apart from noise reduction.
So it would stand to reason that F mount lenses may be less precise and accurate compared to equivalent Z mount lenses - and this may be more obvious with fast lenses.
Thoughts?
I've not seen evidence supporting either of those speculations. The F AF lenses I've tried work great on the Z cameras. When I tested the E version of the 70-200/2.8 against the S version on Z cameras, there was very little to choose between their images. The 180-400/4 is fantastic on the Z.
Neither of those are f1.2-f1.4 lenses though; I am mainly asking about lenses with this aperture range.
Also accuracy tests tend to be for static objects, if I am not mistaken? If the object was moving (not fast) - such as a human model, would there be any noticeable difference in AF performance?
Other F lenses that work great on the Z are the 58/1.4 and the 105/1.4. Dynamic AF on the D5 was very fast, and the most expensive lenses that supported that camera were also very fast. Of course, accuracy is better with a MILC.